New Office hours: Tue 3.30-4.30, CS 187, or by appointment
If you are registered, please join google groups which will be used for class-related announcements and discussions.
Synoposis: This is a fast-paced graduate class on machine learning, covering the foundations, such as (Bayesian) statistics and information theory, as well as supervised learning (classification, regression), and unsupervised learning (clustering, dimensionality reduction, graphical models).
For a slower version of this class, which does not cover unsupervised learning, check out Stat406, also offered in Spring 2010.
Textbook: Draft copies of my textbook, Machine Learning: a probabilistic approach (MLAPA), is available for purchase for $56.50 from Copiesmart in the UBC Village (next to Macdonald's).
Pre-requisites. Linear algebra, calculus, probability theory, programming (preferably Matlab), some undergrad class on machine learning (eg CS 340) or statistics (eg Stat 306).
Grading: Midterm 35%, Homeworks 35%, project 30%
L# | Date | Topic | Reading | Homework |
---|---|---|---|---|
L1 | Tue Jan 5 |
Admin, intro to classif and regr, entrance quiz | 1.1 | hw1.pdf, due Tue 12th. See also Getting started in Matlab for help. |
L2 | Thu Jan 7 |
Knn, CV | 1.4 | . |
L3 | Tue Jan 12 |
Intro to unsupervised | 1.2-1.3, 11.1-11.2 | hw2.pdf, due Tue 19th. |
L4 | Thu Jan 14 |
Ridge, robust regression | 11.3-11.5, robust regression handout | . |
L5 | Tue Jan 19 |
Logistic regression | 12.1-12.4, logistic regression handout | . |
L6 | Thu Jan 21 |
Neural networks | 14.1-14.2 | hw3, due Thur 28th, spamData.mat, |
L7 | Tue Jan 26 |
EM for robust and probit regression; Naive Bayes classifiers | EM for regression and classification handout | . |
L8 | Thu Jan 28 |
Discriminant analysis | Generative classifiers handout, and Fisher's LDA handout | hw4, due Thur 4th |
L9 | Tue Feb 2 |
Missing data in generative classifiers | Mixture models | . |
L10 | Thu Feb 4 |
Mixture models | Regularized discriminant analysis, EM for MAP estimation of mixtures, bankruptcy.txt data, needed for HW5, Fraley05, needed for HW5. | hw5, due Thur 11th |
L11 | Tue Feb 9 |
L1 regularization | Variable selection handout | . |
L12 | Thu Feb 11 |
More on variable selection | . | . |
L13 | Tue Feb 16 |
olympics | . | . |
L14 | Thu Feb 18 |
olympics | . | hw6, due Tue 2nd |
L15 | Tue Feb 23 |
olympics | . | . |
L16 | Thu Feb 25 |
olympics | . | . |
L17 | Tue Mar 2 |
Review session | . | . |
L18 | Thu Mar 4 |
Midterm | . | . |
L19 | Tue Mar 9 |
Sparse kernel machines | Sparse kernel machines handout | . |
L20 | Thu Mar 11 |
Sparse kernel machines, Gaussians | . | hw7, due Tue 23rd. adultCensus.zip data |
L21 | Tue Mar 16 |
Bayesian inference | Bayesian inference for single parameter models | . |
L22 | Thu Mar 18 |
No class | Bayesian stats 1 | . |
L23 | Tue Mar 23 |
More Bayesian stats | Bayesian stats 2 (26mar10 version, slightly updated from 22mar10 version). | . |
L24 | Thu Mar 25 |
Empirical, hierarchical and variational Bayes | Variational inference (26mar10 version, incomplete). See also Bishop's chapter on variational inference | hw8, due Thur 1st. |
L25 | Tue Mar 30 |
More variational Bayes | Handout on conjugate duality and constrained optimization, Handout on variational inference | . |
L26 | Thu Apr 1 |
Graphical models | . | Hw8 due |
L27 | Tue Apr 6 |
HMMs and Kalman filters | Handout on Kalman filters, Handout on HMMs | 1 page project proposal due |
L28 | Thu Apr 8 |
Forwards-backwards algorithm, and friends | . | . |
L29 | Tue Apr 13 |
Monte Carlo inference | Ch 20 of 2jan10 version of book | . |
L30 | Thu Apr 15 |
Inference in graphical models | Handout to appear later | . |
. | Tue Apr 27 |
Project presentations | . | . |